Early in the CMI project, the Storage Group identified existing wells in oil and gas fields as a potentially important leakage pathway, especially in North America. Oil and gas fields offer an appealing opportunity for storage, since their existence proves that a seal can keep hydrocarbons contained for millions of years and much is known about subsurface strata in well-explored areas. However, because oil-producing areas in North America have been punctured by many thousands of existing wells, the seal integrity of well cements is a critical factor in determining whether CO2 will stay stored, or leak up through these potential conduits to the surface. Assessing the leakage potential of existing wells, and its impact on risk assessment, has thus become the primary focus of the Storage Group.


Cement durability experiments

An issue critical to the potential for leakage is the reaction of well cements exposed to carbonated brines that would result from CO2 injection. The group’s early series of flow-through and batch experiments suggest that cements would be highly susceptible to attack from carbonated brines flowing through sandstone reservoirs. In the flow-through experiments, cement was destroyed at a rate on the order of 0.1 mm/week under conditions comparable to sequestration in a sandstone formation at a depth of 1-2 km. More likely rates of attack have been probed using batch experiments, in which the acid must diffuse through the pores of stone to reach the cement. The cement samples exposed to brines in the batch reactors were not as quickly degraded as those in the flow-through experiments, but in one month reaction rings formed at the cement/stone interface caused permeability to increase by an order of magnitude before subsequently leveling off.

In contrast, the team’s experiments showed that the rate of attack was undetectably small when the brine was equilibrated with limestone prior to contact with the cement, because the acidity is reduced and the calcium content of the brine is increased by such exposure. These results imply that sequestration in limestone formations would be much less likely to result in cement corrosion than sequestration in sandstone formations.

Scherer’s group also established collaboration with National Energy Technology Laboratory (NETL) laboratory in Pittsburgh to study cement corrosion under high-pressure conditions that more closely approximate the environment in a storage formation. The NETL group’s initial results indicate that rates of cement corrosion decrease at elevated pressure, which would improve the outlook for carbon storage in deep saline aquifers. Work is ongoing both to quantify this pressure dependence and to identify the mechanism responsible.

To support modeling of the leakage potential of existing wells, additional data is needed regarding the transport and mechanical properties of corroded cement already exposed to carbonated brines, which should provide a more realistic analogue for the conditions likely to be found in aging wells. Previous results indicated a catastrophic loss of integrity of the cements when the corrosion is well advanced. However, there may be a more subtle deterioration that occurs in limestone formations, where our short-term (i.e., 1-year) experiments indicate little or no attack, that might have significant impact on leakage over the course of a century. Therefore, a new graduate student, Ed Matteo, is undertaking experimental studies of the structure and properties of cement following exposure to carbonated brine. Ed is presently using NMR to measure the transport rate of fluids in the degraded cement, to improve our prediction of corrosion rates.


Modeling brine-cement interactions

The data from cement durability experiments are informing simulations of brine-cement interactions in existing wells to assess the overall potential of CO2 leakage from underground aquifers. The likely leakage pathways along existing wells involve micro-annuli or other flow paths that tend to occur at the interface between well cement and rock, or between well cement and casing. These flows can be modified or controlled by geochemical reactions that occur on the scale of the leakage pathways, and inclusion of these important small-scale processes in largescale continues to be a significant computational and conceptual challenge.

In response to this need, the Storage group has been developing a geochemistry module to couple with reactive transport modules in the geomechanical model Dynaflow. The resulting multiphase, multi-component model includes geochemistry appropriate for reaction and degradation of well cements in the presence of carbonated brine solutions, resolves small-scale behavior in the vicinity of wells, and describes the cement reactions as well as phase behavior of the fluid phases. With this model we can now study system behavior and leakage patterns along different kinds of wells, including relevant multi-phase flow physics, flash calculations, and cement reactivity.

The modular nature of the software is particularly important. The module for non-reactive equilibration of the components (i.e., flash calculation) includes all the CO2-rich phases, and is supported by a separate module that calculates all the thermodynamic properties of the system. Bruno Huet has assembled the most extensive database in existence for reactions involving components of phases in cement, and can calculate the composition of the solution in equilibrium with any assemblage of cement phases between 0 and 100˚C. All of these modules are currently integrated by Jean Prévost into Dynaflow, which has exceptional ability to couple geomechanics (including fracture) with geochemistry. However, all of these modules could also be used with other systems, such as Eclipse.

In the last year, enormous progress was made by Bruno Huet and Jean Prévost in developing the reactive transport model to predict the rate of attack of carbonated brine on cement. (Bruno Huet has recently joined Schlumberger in Clamart, France. We hope to continue collaborating with Bruno on the development of the reactive transport code.) The model correctly reproduces the rates of advance of reaction fronts in paste that were measured by Andrew Duguid, now working for Schlumberger in their Carbon Services program in Pittsburgh (Figure 7).

Figure 7. Comparison of thickness of reacted layers measured by A. Duguid and calculated using the reactive transport model developed by B. Huet and J.-H. Prévost.

After combining experimental data with simulation results, the mechanisms of cement reactivity in CO2-saturated brine are now better understood. Our original interpretation of Andrew Duguid’s experiments was that the precipitated layer of calcium carbonate was protective, but the situation is more subtle. The calcium carbonate layer may control the rate of attack initially, causing the linear rate of growth of the corroded layer. However, later in the process the rate of diffusion through the gel layer begins to dominate. The ability of the CaCO3 layer to plug the cement porosity (and thereby delay attack) depends on two things: the solubility of calcite in the aqueous phase (which depends on the CO2 content of the brine) and the rate at which calcium diffuses through the gel layer into the brine. The implications of these mechanisms for the healing of cracks by precipitation of calcite are being examined.

Focus has now turned to the flash calculation, which predicts the phases present under given conditions of temperature, pressure and composition. This is a major challenge when the system crosses a phase boundary, such as boiling of CO2 when carbonated brine rises through a crack. Following widespread consultation among experts in the field (including Lee Chin at Conoco/Phillips and G. Pope’s group at UT Austin), it was agreed that none of the existing codes can handle this situation! Prévost and Lee Chin have recently developed the first flash calculation that does handle this problem correctly, and they are now implementing it in the reactive transport program.

Another major goal of our group is to evaluate the risk of creating leaks in a reservoir as the pressure of injected CO2 deflects the cap rock: since the stiffness of the rock differs from that of the cement plugs that pass through it, cracks might form in the cement that would permit leakage. This question is being investigated by Jean Prévost and a new graduate student, Zhihua Wang, using the poromechanical capabilities of Dynaflow.